eera5607

RAPID is DRAM caching, which is different.

Try using something like crystal disk mark instead as it's way easier to see stuff.
 
It could just be because your drive is 80% full, really.

Does it have lots of airflow?
maybe it's thermal throttling (a real problem with M.2 SSDs)

Ideas:
 
1) It's not plugged into a 6Gb/s sata slot (I did this for years without noticing!) 
2) Thermal throttling.. very unlikely over a sustained period
3) The drive is nearly full. SSDs do not like being full. They slow a LOT as they approach max capacity. Clear your drive as much as possible. 
4) Try Samsung Magician to check on the health of the drive.. it might just be a bad drive/dying. 

Excuse point number 1, I had read that as an SSD not M.2 NVME. 

It's a m.2 drive

I did comment on that right below my original post, tired eyes. My bad.

Oh i see

Try taking off the cover, pointing a fan on it, and running the test again.

The writes seem a bit low... Have you made sure that the SLC caching is enabled in Samsung Magician (they call it something other than SLC caching but yeah)

I posted something similar here: 
Not sure why it's performing like that even with Samsung's own software.
 
 

Actually I just figured out what the SLC caching is called, Samsung calls it turbowrite.

This is what you should expect with the 500GB 960 Evo:
 

 
So your CrystalMark benchmark is right where it should be.

Are you thinking of RAPID or TurboWrite?
 
Never mind.

I've got one as well, but I found that he already found out which one. Did the Samsung driver help your EVO?

Well it seems that it is working as expected now. Now UserBenchmark says the EVO 90 is "Performing way above expectations (91st percentile)" (Complete results: http://www.userbenchmark.com/UserRun/4965874). A big difference with original benchmark: ¨Performing way below expectations (15th percentile)" .

This is what I did:
I discovered that Corsair Link shows the SSD temperature and tested it after placing the GPU on the PCIe x16 slot that is not blocking the air intakes of the PCIe M.2 slot where the SSD is connected to the motherboard. During the benchmarks the SSD reached 39 °C and that seems to be ok. 

I also installed the oficial NVMe drivers from Samsung since it was using the Windows ¨generic¨ drivers. That in combination with changing the GPU position helped the performance on UserBechmark A LOT. Thank you all for the ideas and contribution to solve this issue.

Thanks @DocSwag @RadiatingLight and @Zebedeeboing ! I discover something that help a little. I had the Windows NVM driver installed:

and not the oficial driver that you can download from Samsung's website:

I keep getting a benchmark "below the expectations" but before it was "way below expectations" on UserBenchmark. 
The problem in this benchmark is definitely related to this section:

I still have to try taking out the GPU and taking  off the cover of the SSD to check if it is related to temperatures. The thing is that is working fine with the other two tests and bad with the last one (DQ Read, DQ Write, DQ Mixed).
 
Thanks!
 

Well it seems that it is working as expected now. Now UserBenchmark says the EVO 90 is "Performing way above expectations (91st percentile)" (Complete results: http://www.userbenchmark.com/UserRun/4965874). A big difference with original benchmark: ¨Performing way below expectations (15th percentile)" .

This is what I did:
I discovered that Corsair Link shows the SSD temperature and tested it after placing the GPU on the PCIe x16 slot that is not blocking the air intakes of the PCIe M.2 slot where the SSD is connected to the motherboard. During the benchmarks the SSD reached 39 °C and that seems to be ok. 

I also installed the oficial NVMe drivers from Samsung since it was using the Windows ¨generic¨ drivers. That in combination with changing the GPU position helped the performance on UserBechmark A LOT. Thank you all for the ideas and contribution to solve this issue.

Make sure your Windows is fully updated. But it should work
https://www.microsoft.com/en-us/download/details.aspx?id=50395
 

Apparently PowerShell 5.0 had some issues with Windows 8.1 (not pro). I installed PowerShell 5.1 and everything worked fine. Thank you. 

always charge 2 at a time

Yes, you can.
 
300mA is a perfectly reasonable charge current for modern NiMH batteries. You'll fully charge the batteries in about 6-8 hours.
 

No problem.
niMH batteries are 1.2v nominal, ~ 1.35v when fully charged.
The voltage has to be higher than max voltage to charge so 1.4v per battery (2x1.4=2.8v) is perfectly reasonable.
 
You can learn more about NiMH batteries here: http://data.energizer.com/pdfs/nickelmetalhydride_appman.pdf
 

It's better to install it in a 16x slot but technically you can install it into a 4x slot and it wouldn't be bottle necked. PCIE 3.0 4x is the same as pcie 2.0 8x or pcie 1.0 16x and pcie 1.0 16x doesn't bottle neck any modern gpu

The only problem there is with putting a gpu in the 8x slot is sometimes it has a harder time recognizing it but you should be fine bandwidth wise.